Distributed task execution on grid
This aspect is currently only relevant for CLC Drug Discovery Server.
When executing a distributed task in a grid environment, a number of task executors are scheduled on the grid using the "shared native specification" of the chosen grid preset. Once a task executor has been started on the grid, it will perform computations until the distributed task is completed. Any task executor which has not been run (but has been queued) will be canceled once the distributed task is complete.
The actual number of task executors depends on the "Maximum number of concurrently running task executors per distributed task"-setting. See Distributed task execution for more information about this setting. Most task executors will request only a single thread (i.e. COMMAND_THREAD_MIN and COMMAND_THREAD_MAX are both 1), meaning that the maximum number of concurrently running task executors should be set to the number of cores that should be used to execute tasks on the grid. If the setting is 0, a task executer is queued for each sub-task created by the overall task. It is recommended not to use a value of 0.